P a g e 10 | 16
the road today, where both the longitudinal (braking/accelerating, e.g., adaptive cruise control) and lateral
(steering, e.g., assisted lane-keeping) control of the vehicle is shifting to ADAS. Further enhanced levels
of automated driving functionality include autopilot (Tesla), intellisafe (Volvo), and Distronic plus
steering assist (Mercedes). Overall this fast pace of market penetration of ADAS in vehicles has not
allowed drivers to develop understanding of new systems over an extended period.
The most common taxonomy to capture the development of ADAS technology in cars are SAE’s levels
of automation sae. This approach is based on six levels of automation, ranging from no automation (level
0) to full automation (level 5). In particular, in levels 2/3, the automated system can take partial control
of vehicle, where level 2 expectations of the human driver are to monitor the system and intervene
appropriately, while the level 3 expectation of the human driver is to intervene appropriately upon a
request from the system. Today most ADAS technology equipped cars are at level 1, in which
progression to partial/semi-automation (level 2/3) with in-built ADAS technology in even lower-priced
car models is becoming more common. Also, level 2/3 automation will likely be reality for some time to
come, given that fuller automation (4/5) is emerging slowly without clear market deployment roadmap.
One of main challenges that arise in level 2/3 automation is transition of control from the ADAS to the
human driver, often referred to as the “handover problem.” This transition is, according to social factors
and safety research, a phase where human attention and reliability is critical, but where humans tend to
underperform in those respects son2017situation. E.g., research has indicated that automatic cruise
control technology leads to a reduction in mental workload and, thus, to problems with regaining control
of the vehicle in failure scenarios stanton1998vehicle. Additionally, a common misconception
concerning ADAS technology is that when more automation is introduced, human error will
disappear atlantic2015save, which may give rise to the problematic idea that driver training is not
necessarily needed. However, social factors research advises against not training for the use of new
sophisticated automation technology lee2006human; salas2006design; saetren2015effects, as humans in
the technology loop will still be needed for use, maintenance or design of the technology. It may even be
that increased automation increases the level of competence required for the driver, as the driver must
know both how to handle system manually, for instance, if the sensors in a car stop working due to bad
weather, in addition to knowing how to control and supervise the advanced automation technology.
In our previous work rismani2018qualitative, we performed a qualitative survey and found that the
handover problem is challenging, and it is unclear to drivers how this could best be handled securely.
Furthermore, drivers were worried about the implications of vehicle automation due to lack of knowledge
and experience of level 2/3 systems and seemed concerned about the kind of training and licensing that
accompanies these developments in vehicle automation. The lack of certainty around training and
licensing concerning emerging ADAS technologies is a relevant ethical concern, as it exposes a gap in
regulation and industry best practices that have not been the focus of much research to date.